This project enhances the “FlightPath” mixed reality app for rapid neurocognitive impairment (NCI) screening by developing a reinforcement learning (RL)-based algorithm to simulate realistic, unpredictable hummingbird flight patterns. These adaptive trajectories generate enriched user interaction data, improving machine learning-based detection of conditions like concussion and dementia. The RL model mimics complex maneuvers—hovering, sharp turns, and evasive actions—tailored to user performance. Integrated into FlightPath, this innovation supports clinician-AI collaboration for standardized, non-invasive assessment. In partnership with UIUC’s Jump ARCHES team, the project lays groundwork for future NIH/NSF funding and clinical translation of MR-based cognitive evaluation tools.
Echocardiography is essential to cardiovascular care but remains limited by operator variability, inconsistent image quality, and time-intensive manual measurements. This project develops a generative-AI pipeline to enhance ultrasound images, automate cardiac chamber segmentation and ejection-fraction measurement, and provide explainable decision support while retaining clinician oversight. Models will be trained and validated against expert tracings using OSF HealthCare datasets and evaluated in a pilot clinical-review interface. Expected impact: faster and more consistent echo interpretation, reduced manual burden for clinicians, improved training resources, and groundwork for future NIH/NSF funding and clinical translation.
Opioid overdose remains a leading cause of preventable death in Illinois, yet access to naloxone training is uneven, especially for urban and campus populations. Phase I demonstrated that the Virtual Extended Naloxone Training (VeNT) platform improves knowledge and confidence in rural communities. Phase II expands VeNT to address urban equity barriers by adding Spanish-language content and testing the platform with Illinois State University students and community members. A randomized controlled trial will compare VeNT to traditional training using validated overdose-knowledge and self-efficacy measures, plus in-session performance tasks and community naloxone-distribution data. Results will inform statewide, scalable, culturally responsive overdose-prevention deployment.